fix(swarm-coordination-registry): [#2226] recompute inactivity cutoff per tick

[josecelano]

Summary

The activity metrics updater now recomputes the peer inactivity cutoff before every update tick
instead of reusing a timestamp captured at job startup.

Problem

run_job computed now - max_peer_timeout once and passed that absolute timestamp to every
tick. Any peer announced after the tracker started therefore had updated > cutoff for the rest
of the process lifetime and was never reported inactive: the
swarm_coordination_registry_peers_inactive_total and ..._torrents_inactive_total gauges
stayed at 0. Peer cleanup has its own job and was unaffected; only the metrics were wrong.

Changes

• run_job takes max_peer_timeout (the configured duration) instead of a precomputed cutoff;
  bootstrap wiring updated accordingly.
• The cutoff derivation is extracted into a pure inactivity_cutoff(now, max_peer_timeout)
  and called with CurrentClock::now() on every tick.
• Doc comments on Peer::updated and the updater clarify that the cutoff is an absolute
  timestamp; the naming question around DurationSinceUnixEpoch is parked as a follow-up draft
  in the issue folder.

Tests

• activity_metrics_updater::tests::inactivity_cutoff: three pure unit tests (timeout before
  now; zero timeout yields now; now earlier than timeout falls back to the epoch so nothing is
  reported inactive).
• statistics::tests: collaboration tests running the job against a real Registry and
  Repository through a state-named fixture (JobWithOnePeerAnnouncedAtStartup), with Tokio
  paused time driving the interval and clock::Stopped driving domain time:
  • peer active 1 s after startup with a 2 s timeout;
  • peer inactive 3 s after startup;
  • cutoff recomputed on every update: same job and peer observed active on the first tick and
    inactive on the second.
• Mutation checks recorded in the commit history: reintroducing the startup-captured cutoff
  fails the after-timeout test; caching the cutoff after the first tick is caught only by the
  two-tick test. Production code was mutated in the working tree and restored, never committed.
• Missing tests identified during the behaviour inventory are noted in the owning modules and
  tracked by Overhaul: Packages Testing #1347.

Documentation riding along

Lessons from this fix, made repository-owned:

• write-unit-test skill: inventory what a module decides vs. forwards and choose the test
  level; two independent clocks in async job tests; prove a regression test fails against the
  reintroduced bug.
• New catalog entry: pure decision function for collaborator-observed outcomes.
• AGENTS.md: keep intermediate commits; minor process fixes may ride along in the PR that
  produced them.

Validation

cargo test -p torrust-tracker-swarm-coordination-registry --lib statistics: 30 passed.
Pre-commit (8 checks) and pre-push (nightly fmt, nightly check, nightly doc, full stable test
suite) pass on the rebased branch. Runtime verification against a fixed build is recorded in the
issue folder (manual-verification-evidence.md).

Closes #2226
Related to #1347

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Pull request overview

This PR fixes the swarm coordination registry activity metrics job so the “inactive peer” cutoff is recomputed on every update tick (instead of using a stale startup cutoff), and adds a regression test using a stopped clock to validate the behavior.

Changes:

• Pass max_peer_timeout into the registry metrics job and compute the inactivity cutoff inside each tick.
• Remove the bootstrap-layer helper that computed a single cutoff timestamp at startup.
• Add a stopped-clock regression test to ensure peers become inactive after time advances post-start.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/bootstrap/jobs/activity_metrics_updater.rs	Adjusts bootstrap wiring to pass max_peer_timeout instead of a precomputed cutoff.
packages/swarm-coordination-registry/src/statistics/activity_metrics_updater.rs	Recomputes inactivity cutoff per tick and adds a regression test using a stopped clock.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.79%. Comparing base (2b72ecd) to head (e35a1a6).

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2252      +/-   ##
===========================================
+ Coverage    85.49%   85.79%   +0.29%     
===========================================
  Files          353      353              
  Lines        31333    31418      +85     
  Branches     31333    31418      +85     
===========================================
+ Hits         26789    26954     +165     
+ Misses        4165     4087      -78     
+ Partials       379      377       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

🟡 Changes recommended

The bootstrap job file no longer imports std::future::Future while still using -> impl Future<...>, which will fail compilation.

Get a fresh assessment by requesting another Copilot review.

Review details

• Files reviewed: 6/6 changed files
• Comments generated: 2
• Review effort level: Lite

[Copilot]

🟡 Changes recommended

The new paused-time collaboration tests and the updated write-unit-test guidance assume a current-thread Tokio runtime for thread-local clock::Stopped, but the tests/docs don’t consistently enforce or state that, risking flaky/incorrect results.

Get a fresh assessment by requesting another Copilot review.

Review details

Suppressed comments (2)

packages/swarm-coordination-registry/src/statistics/mod.rs:238

• Same determinism issue as the earlier collaboration test: because the job is tokio::spawned and clock::Stopped::local_set is thread-local, this test should force a single-thread Tokio runtime to ensure the spawned job observes the same stopped-clock timeline.

    #[tokio::test(start_paused = true)]

packages/swarm-coordination-registry/src/statistics/mod.rs:257

• Same determinism issue as the other collaboration tests in this module: without forcing flavor = "current_thread", the tokio::spawned job can run on a different worker thread and miss the test’s clock::Stopped::local_set updates, making the two-tick regression guard flaky.

    #[tokio::test(start_paused = true)]

• Files reviewed: 17/17 changed files
• Comments generated: 2
• Review effort level: Lite

[josecelano]

ACK e35a1a6